Prediction Model Of New Onset Atrial Fibrillation In Patients With Acute Coronary Syndrome

BackgroundAcute coronary syndrome(ACS)is a life-threatening heart condition associated with a sudden reduced blood flow to the heart.Atrial fibrillation(AF),which is the most clinically significant cardiac arrhythmia,often manifests as the most common complications of ACS patients.The incidence of new onset AF(NOAF)in ACS patients reported in China is between 6.7% and 13.4%.Studies have shown that ACS patients with NOAF during hospitalization have a higher risk of serious adverse outcomes than patients without AF.In recent years,clinical epidemiological studies have found possible risk factors related to NOAF in ACS patients,and established prediction models based on a few clinical features or single biomarkers.However,there were some limitations in these prediction models,which included: low discrimination of models,lack of comprehensive evaluation of models,lack of external validation of models,etc.In addition,ACS has three clinical subtypes that shall be distinguished by their risk factors of onset and prognosis.Unstable angina(UA)is the most common subtype of ACS.The related risk factors of NOAF in patients with UA may be different from patients with other ACS subtypes.Therefore,it is important to carry out research on the prediction models of NOAF in ACS patients in general,as well as in patient with UA subtype in particular,so as to provide scientific basis for the early detection system of high-risk ACS patients.ObjectivesTo screen the related risk factors of NOAF in patients with ACS and UA during hospitalization,build a prediction model,validate and evaluate the model internally and externally,and further visualize the model,so as to assist clinical decision-making by providing a scientific basis for convenient and fast quantitative assessment of ACS patients’ risk of NOAF.MethodsThe subjects were from cardiology department of two Grade III Level A hospitals,eligible ACS and UA patients were continuously enrolled and divided into modeling group and external validation group.The demographic characteristics,physical examination results at admission to hospital,laboratory test indicators,echocardiography results and treatment regimens were collected retrospectively.The clinical characteristics of patients were described in modeling group and external validation group.The model group used multi-factor Logistic regression and Lasso regression to screen the NOAF predictors.The prediction models were then constructed based on Logistic regression equation.The model was validated in the external validation group.The discrimination of the models were assessed by ROC curve and area under the curve(AUC).The calibration curve was drawn to evaluate the model calibration degree.The clinical utility of the model was evaluated by decision curve analysis.The nomograms were drawn separately,and interactive web pages were made to visualize the prediction model.Results1.A total of 1,535 ACS patients were included in the modeling group,including 820 cases of UA;1,635 ACS patients were included in the external validation group,including1,288 cases of UA.During hospitalization,the incidence of NOAF was 8.21% in the modeling group and 6.12% in the external validation group.In both the modeling group and the external validation group,the patients with NOAF had older age,higher admission heart rate,higher BNP level and larger left / right atrial diameter compared with the patients without AF.2.In the modeling group of patients with ACS,through multivariate Logistic regression and Lasso regression analysis,we found that age,admission heart rate,left atrial diameter,right atrial diameter,heart failure grade,BNP level,statins,and PCI treatment can be used as independent predictors of NOAF.Based on these indicators,the prediction model was constructed and verified.The model evaluation results showed that the AUC of modeling group was 0.891(95% CI: 0.863-0.920),AUC of validation group was 0.839(95% CI:0.796-0.883),and both groups passed the calibration test(P > 0.05).The clinical utility evaluation shows that the model has a clinical net benefit within a certain range of the threshold probability.3.In the modeling group of patients with UA,multivariate Logistic regression and Lasso regression analysis showed that age,left atrial diameter,right atrial diameter,heart failure grade,BNP level and PCI treatment could be used as predictors of NOAF in patients with UA.The AUC of modeling group was 0.894(95% CI: 0.854-0.934),and that of validation group was 0.844(95% CI: 0.790-891).At the same time,the model possessed a good degree of calibration and a clinical net benefit.4.The models were visualized by nomograms and interactive web pages,which could quantitatively and visually demonstrate the risk of NOAF.Conclusions1.Older age,higher heart rate on admission,increased left and / or right atrial diameter,increased BNP-a biomarker of cardiac function,and heart failure were independent risk factors for NOAF in ACS patients.At the same time,statins and PCI were protective factors for NOAF.2.Older age,increased left and / or right atrial diameter,increased BNP,and heart failure were also independent risk factors for NOAF in patients with UA.PCI was the protective factor for NOAF.3.The prediction models of NOAF in patients with ACS and patients with UA were constructed,respectively.The internal and external validation and evaluation of the models confirmed that the models had good discrimination,stable calibration and certain clinical net benefit.The visualization of prediction model makes it easier for medical workers to quantitatively assess the risk of NOAF in early stage,and provides important scientific basis for clinical decision-making.